ST 1.0.9 problem - all files ignored

[marekl]

Hi All,

I have tried to train acoustic model using "long audio alignment" training
(SphinxTrain 1.0.9). Model was successfully trained on one of earlier versions
of ST. Unfortunately ST 1.0.9 ignores all files in both LDA and not LDA
training modes due to align problems:

INFO: cvt2triphone.c(199): no multiphones defined, no conversion done
    90ERROR: "forward.cu", line 740: All 49 active states, 10 11 12 13 14 140 141 142 143 144 20 21 22 23 24 125 126 127 128 129 50 51 52 53 54 90 91 92 93 94 50 51 52 53 54 10 11 12 13 14 170 N(2,5) N(28,5) N(4,5) N(25,5) N(10,5) N(18,5) N(10,5) N(2,5), zero at time 8
ERROR: "backward.c", line 456: Failed to align audio to trancript: final state of the search is not reached
ERROR: "baum_welch.c", line 337: Apartus/Apartus10 ignored

Suggestions from similar topic, discussed earlier already checked with no
result.

Any ideas?

Thanks in advance

[Nickolay V. Shmyrev]

Hello Marek

This branch hasn't not been merged, it's not 1.0.9, it's just a long audio
trainer branch. It's work in progress now, not supposed to be working or
tested.

[marekl]

Hello Nicolay,

Nice to hear from you again.

Could you tell me when this branch is supposed to reach beta-test phase. May I
be helpful with testing or "polishing" this branch?